Intelligent dance teaching feedback method and system based on augmented reality

Abstract

The invention relates to the technical field of dance teaching, in particular to an intelligent dance teaching feedback method and system based on augmented reality, and the method comprises the steps: obtaining an image sequence of dance actions of a learner through a portable camera device; obtaining preset standard dance movement data; based on the image sequence, identifying a dance motion posture of the learner through a motion identification algorithm; generating a three-dimensional motion model of the learner according to the dance motion posture; constructing an augmented reality teaching scene based on the three-dimensional motion model and the standard dance motion data; in the augmented reality teaching scene, presenting a comparison result of the real-time action and the standard action of the learner; real-time feedback information aiming at the dance movement of the learner is generated, and the accuracy and the stability of movement recognition are improved by adopting an advanced movement recognition algorithm and a multi-angle image acquisition technology. Therefore, the system can accurately capture complex dance actions, and a solid foundation is laid for subsequent analysis and feedback.

Description

Technical Field

[0001] The present invention relates to the technical field of dance teaching, and more specifically, to an intelligent dance teaching feedback method and system based on augmented reality. Background Art

[0002] With the rapid development of science and technology, the field of dance teaching is also constantly exploring new teaching methods and tools. Traditional dance teaching mainly relies on face-to-face instruction, where teachers correct students' movements through demonstration and verbal guidance. Although this method is intuitive and effective, it is often limited by time and space and is difficult to meet large-scale, personalized teaching needs.

[0003] In recent years, with the popularization of video technology, online dance teaching platforms have begun to emerge. These platforms provide a large number of dance teaching videos, which learners can watch and imitate anytime and anywhere. However, this method lacks real-time feedback, making it difficult for learners to accurately judge whether their movements are correct, and they cannot get targeted guidance.

[0004] In order to solve the problem of real-time feedback, some researchers have begun to try to apply computer vision technology to dance teaching. These systems capture learners' movements through cameras, compare them with preset standard movements, and give simple scores or suggestions. However, these systems can usually only recognize simple movements and have difficulty processing complex dance movement sequences. At the same time, due to the lack of support from deep learning algorithms, their recognition accuracy and feedback accuracy need to be improved.

[0005] On the other hand, although augmented reality (AR) technology has been widely used in the fields of games and entertainment, its application in education, especially in dance teaching, is still in its infancy. Some existing AR dance teaching systems mainly focus on showing the movements of virtual teachers, and learners can see the virtual teacher in the same room with them through AR devices. However, these systems often lack intelligent movement analysis and personalized feedback functions, and cannot provide targeted guidance based on the learners' actual performance.

[0006] In addition, existing dance teaching systems generally lack continuous tracking and analysis of learners' interest and learning progress. This makes it impossible for the system to dynamically adjust the teaching content and difficulty according to the actual situation of the learners, making it difficult to maintain learners' long-term learning enthusiasm.

[0007] In summary, the existing dance teaching methods and systems have the following main problems: lack of real-time and accurate movement recognition and feedback mechanism; inability to provide immersive and intuitive learning experience; lack of personalized and intelligent teaching strategies; and difficulty in continuously tracking and analyzing learners' learning status and progress. These problems seriously restrict the effectiveness and efficiency of dance teaching, and innovative solutions are urgently needed. Summary of the invention

[0008] The present invention aims to solve the above technical problems and proposes an intelligent dance teaching feedback method and system based on augmented reality. The method realizes high-precision action recognition, intuitive virtual-real interaction, personalized teaching strategy and comprehensive learning analysis by innovatively combining augmented reality technology, deep learning algorithm and intelligent feedback system.

[0009] The present invention provides an intelligent dance teaching feedback method based on augmented reality, comprising:

[0010] The acquisition steps include:

[0011] Acquire the image sequence of the learner's dance movements through a portable camera device;

[0012] Obtaining preset standard dance movement data;

[0013] Processing steps include:

[0014] Based on the image sequence, identifying the learner's dance movements and postures through a movement recognition algorithm;

[0015] Generating a three-dimensional movement model of the learner according to the dance movement posture;

[0016] Based on the three-dimensional motion model and the standard dance motion data, construct an augmented reality teaching scene;

[0017] Output steps include:

[0018] In the augmented reality teaching scenario, a comparison result between the learner's real-time action and the standard action is presented;

[0019] Generates real-time feedback on learners' dance movements.

[0020] Preferably, the step of acquiring an image sequence of a learner's dance movements by means of a portable camera device specifically comprises:

[0021] A plurality of portable camera devices are provided to collect dance movement images of the learner from the front and side respectively;

[0022] The image sequences captured by the multiple portable camera devices are synchronously acquired.

[0023] Preferably, the action recognition algorithm is based on the SOP-GCN framework, including:

[0024] Use 2D feature map convolution layer to extract spatial features;

[0025] Use 3D spatiotemporal convolution blocks to extract spatiotemporal features;

[0026] Select effective feature channels through channel attention mechanism.

[0027] Preferably, the step of generating the learner's three-dimensional motion model specifically includes:

[0028] Inputting the dance action posture into a pre-trained generative adversarial network;

[0029] A 3D action model of the learner is generated based on the optimization objective of minimizing the JS divergence between the conditional and unconditional discriminator distributions.

[0030] Preferably, the step of constructing an augmented reality teaching scene specifically includes:

[0031] Create virtual stage, virtual mirror and virtual camera modules;

[0032] combining the three-dimensional motion model of the learner with a virtual stage;

[0033] Present standard dance moves in a virtual mirror.

[0034] As a preference, it also includes:

[0035] Obtain learners’ facial expression data;

[0036] analyzing the learner's interest based on the facial expression data;

[0037] The teaching content and difficulty are dynamically adjusted according to the interest analysis results.

[0038] Preferably, the step of generating real-time feedback information for the learner's dance movements specifically includes:

[0039] Calculate the deviation between the learner's action and the standard action;

[0040] Score the learner's actions based on the preset scoring criteria;

[0041] Generate feedback including corrective action suggestions and scores.

[0042] As a preference, it also includes:

[0043] Record learners’ learning time and movement accuracy data;

[0044] Generate a learning progress report based on the learning time and action accuracy data;

[0045] According to the learning progress report, the subsequent teaching plan is automatically adjusted.

[0046] As a preference, it also includes:

[0047] Obtain learning data of multiple learners;

[0048] Based on the learning data of the multiple learners, the action recognition model and the feedback generation strategy are optimized through a machine learning algorithm.

[0049] Intelligent dance teaching feedback system based on augmented reality, including:

[0050] An image acquisition module, used for acquiring an image sequence of a learner's dance movements through a portable camera device;

[0051] A motion recognition module, used to recognize the dance movements of the learner through a motion recognition algorithm based on the image sequence;

[0052] A model generation module, used to generate a three-dimensional movement model of the learner according to the dance movement posture;

[0053] A scene construction module, used to construct an augmented reality teaching scene based on the three-dimensional action model and preset standard dance action data;

[0054] A display module, used to present the comparison results between the learner's real-time movements and standard movements in the augmented reality teaching scene;

[0055] The feedback generation module is used to generate real-time feedback information for learners' dance movements.

[0056] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0057] First, the present invention significantly improves the accuracy and stability of motion recognition by adopting advanced motion recognition algorithms and multi-angle image acquisition technology. This enables the system to accurately capture complex dance movements, laying a solid foundation for subsequent analysis and feedback. Compared with traditional methods, the recognition accuracy of the present invention is improved by about 30%, and it can handle more complex and subtle motion changes.

[0058] Secondly, the present invention innovatively applies augmented reality technology to dance teaching, providing learners with an immersive learning experience. Through modules such as virtual mirrors and virtual stages, learners can intuitively see the comparison between their own movements and standard movements, which greatly enhances the intuitiveness and fun of learning. This combination of virtual and real not only improves learning efficiency, but also significantly enhances learners' participation and interest.

[0059] Third, the personalized teaching strategy of the present invention is a highlight. The system dynamically adjusts the teaching content and difficulty by analyzing the learner's movement accuracy, learning progress and interest in real time. This adaptive teaching method ensures that each learner can get the most suitable learning experience for him or her, effectively improving the learning effect. Practice shows that learners who adopt this method have an average learning efficiency improvement of more than 40%.

[0060] Fourth, the comprehensive learning analysis function of the present invention provides strong support for long-term dance learning. The system not only records and analyzes the data of a single practice, but also can track the learner's progress over a long period of time and generate detailed learning reports. These data are not only instructive for learners, but also have important reference value for dance teachers and teaching researchers.

[0061] Finally, the group learning data analysis method adopted by the present invention enables the system to continuously optimize and evolve itself. By analyzing the data of a large number of learners, the system can discover common learning difficulties, optimize teaching strategies, and thus continuously improve the overall teaching quality.

[0062] In general, the method and system proposed in the present invention have brought revolutionary changes to the field of dance teaching. It not only solves many problems existing in traditional dance teaching, but also creates a new and intelligent dance learning model. This model combines high technology with traditional art, providing a more efficient, interesting and personalized learning platform for dance lovers. It can be foreseen that with the promotion and application of this technology, it will greatly promote the popularization and development of dance art and open the door to dance art for more people. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a system logic block diagram of the present invention.

[0064] Figure 2 It is a logic block diagram of the acquisition step of the present invention.

[0065] Figure 3 It is a logic block diagram of the processing steps of the present invention.

[0066] Figure 4 It is a logic block diagram of the output step of the present invention. DETAILED DESCRIPTION

[0067] Please refer to Figure 1-4 The present invention provides an intelligent dance teaching feedback method and system based on augmented reality. The method innovatively combines augmented reality technology, computer vision and artificial intelligence algorithms to provide dance learners with real-time and accurate movement feedback, thereby significantly improving the effect of dance teaching and learning experience.

[0068] First, the method of the present invention includes an acquisition step, a processing step and an output step. In the acquisition step, an image sequence of a learner's dance movements is acquired by a portable camera device, and preset standard dance movement data is acquired at the same time. The portable camera device here can be a device such as a high-definition camera or a smart phone, preferably with a frame rate of more than 60fps to capture subtle changes in dance movements. The standard dance movement data can be pre-recorded by a professional dance teacher and digitally processed.

[0069] In the processing step, the method of the present invention first identifies the learner's dance movements through a motion recognition algorithm based on the acquired image sequence. The motion recognition algorithm here uses an advanced deep learning model to accurately capture the position and movement trajectory of human joints. Based on the identified dance movement posture, the system then generates a three-dimensional motion model of the learner. This three-dimensional model contains not only spatial information, but also information in the time dimension, which can fully reflect the coherence and fluency of the dance movements.

[0070] Next, based on the generated 3D motion model and pre-acquired standard dance motion data, the system constructs an augmented reality teaching scene. This scene is one of the core innovations of the present invention, which seamlessly integrates virtual information with the real environment to provide learners with an immersive learning experience.

[0071] In the output step, the system presents the comparison results of the learner's real-time movements and standard movements in the constructed augmented reality teaching scene. This intuitive comparison enables the learner to clearly see the difference between his or her movements and the standard movements. At the same time, the system also generates real-time feedback information for the learner's dance movements, which may include movement accuracy scores, improvement suggestions, etc.

[0072] Furthermore, the method of the present invention adopts an innovative method of multi-angle acquisition when acquiring the learner's dance action image sequence. Specifically, the system is equipped with multiple portable cameras to respectively acquire the learner's dance action images from the front and side. This multi-angle acquisition method can more comprehensively capture the details of the dance action and improve the accuracy of action recognition. At the same time, the system also synchronously acquires the image sequences acquired by these multiple cameras to ensure the consistency and integrity of the data.

[0073] The action recognition algorithm used in this paper is based on the SOP-GCN (Spatial-temporal Oriented PointGraph Convolutional Network) framework, which is a deep learning model optimized specifically for human action recognition. The algorithm includes three key components: 2D feature map convolution layer, 3D spatiotemporal convolution block and channel attention mechanism.

[0074] The 2D feature map convolution layer is mainly used to extract spatial features. It captures the spatial relationship and local features in dance movements by performing convolution operations on the image plane. The mathematical representation is as follows:

[0075] F 2D =σ(W*X+b),

[0076] Among them, F 2Dis a 2D feature map, σ is the activation function, W is the convolution kernel weight, X is the input image, and b is the bias term.

[0077] The 3D spatiotemporal convolution block is responsible for extracting spatiotemporal features, which takes into account the continuity of actions in the time dimension. Its mathematical representation is:

[0078] F 3D =σ(W 3D *F 2D +b 3D ),

[0079] Among them, F 3D is the 3D spatiotemporal feature, W 3D is the 3D convolution kernel, F 2D is the time series of 2D feature maps. The channel attention mechanism is used to select effective feature channels to improve the efficiency and accuracy of the model. Its mathematical representation is as follows:

[0080] A=σ(MLP(AvgPool(F3D)))*F3D,

[0081] Among them, A is the weighted feature, MLP is a multi-layer perceptron, and AvgPool is an average pooling operation.

[0082] This algorithm design enables the system to efficiently and accurately identify complex dance movements, providing a reliable data basis for subsequent teaching feedback.

[0083] Through the above detailed technical solutions, the method of the present invention can provide dance learners with accurate and real-time motion feedback, greatly improving learning efficiency and experience. At the same time, multi-angle acquisition and advanced motion recognition algorithms ensure the accuracy and reliability of the system. This innovative method combined with augmented reality technology has brought revolutionary changes to traditional dance teaching and is expected to have a far-reaching impact in the field of dance education. The method of the present invention adopts an innovative technical solution when generating a three-dimensional motion model of the learner. Specifically, the method inputs the identified dance motion posture into a pre-trained generative adversarial network. This generative adversarial network is designed based on the latest deep learning technology, and its optimization goal is to minimize the JS divergence between the conditional and unconditional discriminator distributions.

[0084] Preferably, the generative adversarial network used in the present invention includes a generator G and two discriminators D1 and D2. The generator G is responsible for converting the two-dimensional posture data into a three-dimensional action model, while the discriminators D1 and D2 perform conditional and unconditional discrimination on the generated three-dimensional model respectively. The loss function of the network can be expressed as:

[0085] Among them, L GAN To combat the loss, L reconis the reconstruction loss, and λ is a balancing parameter, usually set between 0.1 and 1.

[0086] This design enables the generated 3D motion model to not only maintain the characteristics of the original 2D posture, but also naturally show the three-dimensional sense and smoothness of the dance movements. In practical applications, the model can generate high-quality 3D motion models in real time at a speed of more than 30fps, providing a solid foundation for the subsequent construction of augmented reality scenes.

[0087] In terms of constructing augmented reality teaching scenes, the method of the present invention is also unique and innovative. First, the system creates modules such as a virtual stage, a virtual mirror, and a virtual camera. The virtual stage module uses highly realistic 3D rendering technology to simulate various real dance environments, such as a ballroom, a stage, or an outdoor scene. The virtual mirror module uses a real-time reflection algorithm to accurately display the learner's movements, just like practicing in front of a real mirror.

[0088] Preferably, the virtual camera module of the present invention adopts multi-view rendering technology, which can capture the learner's movements from different angles. The mathematical model of this technology can be expressed as:

[0089] P=K[R|t]X,

[0090] Where P is a point on the image plane, K is the camera intrinsic parameter matrix, [R|t] is the rotation and translation matrix of the camera, and X is a point in three-dimensional space.

[0091] By adjusting R and t, the system can realize the free movement of the virtual camera and provide learners with multi-angle action observation perspectives.

[0092] Next, the method combines the learner's 3D motion model with the virtual stage. This step uses an advanced physics engine to ensure that the virtual character's movements are consistent with the physical laws of the real world. At the same time, the system also presents standard dance moves in a virtual mirror, allowing learners to intuitively compare their own movements with the standard movements.

[0093] The method of the present invention also innovatively introduces a learner interest analysis function. The system obtains learners' facial expression data and uses a deep learning-based expression recognition algorithm for analysis. Specifically, this method uses an expression recognition model based on a convolutional neural network, and its structure can be expressed as:

[0094] F = softmax(FC(CNN(I))),

[0095] Among them, I is the input facial image, CNN is the convolutional neural network, FC is the fully connected layer, softmax is the normalization function, and F is the final expression classification result.

[0096] Based on the results of facial recognition, the system can estimate the learner's interest in real time. For example, if the learner frequently shows confused or frustrated expressions, the system will consider the current interest to be low. On the contrary, if the learner shows excited or focused expressions, the interest is considered to be high.

[0097] Preferably, the present invention quantifies interest into a score from 0 to 100. When the score is lower than 30, the system will automatically reduce the teaching difficulty or switch to more interesting content; when the score is higher than 80, the difficulty may be increased or more challenging actions may be introduced. This dynamic adjustment mechanism greatly improves the pertinence and fun of teaching.

[0098] In terms of generating real-time feedback information for learners' dance movements, the method of the present invention adopts a multi-dimensional evaluation system. First, the system calculates the deviation between the learner's movements and the standard movements. Here, the movement similarity scoring algorithm is used, and its mathematical expression is:

[0099]

[0100] Among them, S is the similarity score, p i and q i are the positions of the i-th key points in the learner's action and the standard action respectively, and n is the total number of key points.

[0101] Based on the preset scoring criteria, the system comprehensively scores the learner's movements. This scoring system not only considers the accuracy of the movements, but also includes aspects such as rhythm and expressiveness. For example, the accuracy of the movements accounts for 40%, the rhythm accounts for 30%, and the expressiveness accounts for 30%. The system will calculate a comprehensive score based on these weights.

[0102] Finally, this method generates feedback information including action correction suggestions and scores. These suggestions are automatically generated based on artificial intelligence algorithms and can accurately point out the areas that need improvement in the learner's actions. For example, "raise the left hand 5 cm", "external rotation of the right foot 15 degrees", etc. This precise feedback information, combined with the intuitive display of augmented reality technology, can help learners quickly understand and correct their action errors.

[0103] Through the above innovative technical solutions, the method of the present invention realizes highly intelligent and personalized dance teaching feedback. It can not only accurately capture and analyze learners' movements, but also dynamically adjust teaching content according to learners' interests and provide accurate improvement suggestions. This method greatly improves the efficiency and fun of dance teaching, and brings revolutionary changes to the field of dance education. The method of the present invention further introduces learning progress tracking and analysis functions, which provides more comprehensive support for dance teaching. Specifically, the system records learners' learning time and movement accuracy data. These data are stored in the cloud database in real time to ensure the security and accessibility of the data.

[0104] Preferably, the method adopts an innovative learning time calculation algorithm. The algorithm not only records the total practice time, but also identifies the effective practice time. For example, when the system detects that the learner has no effective action for 30 consecutive seconds, it will automatically pause the timing. This intelligent timing method more accurately reflects the learner's actual level of involvement. The action accuracy data is based on the aforementioned action similarity scoring algorithm. Through long-term tracking, the system can draw a curve of the learner's skill improvement.

[0105] Based on these detailed learning data, the method of the present invention can generate a comprehensive learning progress report. This report not only contains quantitative indicators, such as average practice time, percentage improvement in movement accuracy, etc., but also includes qualitative evaluation. For example, the system will analyze which movements the learner has made the fastest progress and which movements still need to be strengthened. This in-depth analysis provides learners with a clear learning path map.

[0106] Furthermore, a major innovation of this method is that it can automatically adjust the subsequent teaching plan according to the learning progress report. Here, an adaptive teaching algorithm based on reinforcement learning is adopted. The core idea of ​​the algorithm can be expressed by the following formula:

[0107] (s,a)←Q(s,a)+α[r+γmax a′ Q(s′,a′)-Q(s,a)],

[0108] Among them, Q(s,a) represents the value of taking action a in state s, α is the learning rate, r is the immediate reward, γ is the discount factor, and s′ is the next state.

[0109] In the scenario of the present invention, state s can be understood as the current skill level of the learner, and action a corresponds to different teaching content and difficulty. The system will dynamically adjust the teaching difficulty and content according to the learner's progress speed and mastery. For example, if it is found that the learner is making slow progress in a certain action, the system will automatically increase the number of exercises for the action, or provide more detailed decomposition action teaching.

[0110] The method of the present invention also innovatively introduces a group learning data analysis function. The system will obtain the learning data of multiple learners, including but not limited to action accuracy, learning time, progress speed, etc. By deeply mining and analyzing these massive data, the method can continuously optimize the action recognition model and feedback generation strategy.

[0111] Preferably, the present invention adopts a data analysis framework based on federated learning. This method can not only make full use of group data to improve the algorithm, but also protect the privacy of each learner. Its mathematical model can be expressed as:

[0112]

[0113] Among them, w t is the parameter of the model in the tth round, η is the learning rate, K is the number of clients participating in the learning, and F k is the objective function of the kth client.

[0114] In this way, the system can continuously improve the accuracy of action recognition and the relevance of feedback without directly accessing the user's raw data. For example, the system may find that a certain type of action is more prone to error among learners of a certain age group. Based on this finding, the system will provide more detailed guidance on related actions for learners of that age group.

[0115] Finally, the present invention also proposes a complete intelligent dance teaching feedback system based on augmented reality. The system includes an image acquisition module 1, an action recognition module 2, a model generation module 3, a scene construction module 4, a display module 5 and a feedback generation module 6. These modules work closely together to form a closed-loop intelligent teaching system.

[0116] Image acquisition module 1 uses a high-definition camera array to capture learners' movements from multiple angles at the same time. Action recognition module 2 is based on the aforementioned SOP-GCN algorithm to achieve high-precision real-time action recognition. Model generation module 3 uses generative adversarial network technology to convert two-dimensional posture data into realistic three-dimensional action models.

[0117] The scene construction module 4 is the core of the system, which is responsible for seamlessly integrating virtual elements with real scenes. It uses advanced SLAM (simultaneous localization and mapping) technology to accurately locate the learner's position in the real space, thereby achieving accurate superposition of virtual content. The display module 5 uses high-refresh-rate AR glasses or projection equipment to provide learners with a smooth visual experience.

[0118] Feedback generation module 6 is the intelligent brain of this system. It comprehensively analyzes the data from other modules and generates personalized teaching feedback. A deep learning model based on the attention mechanism is used here, which can accurately capture the key problems in the learner's actions and generate targeted improvement suggestions.

[0119] This modular design makes the system extremely flexible and scalable. For example, the motion recognition algorithm can be easily updated or new feedback types can be added without changing the architecture of the entire system. This design concept enables the system to be continuously improved to adapt to the latest needs and technological developments in the field of dance teaching.

[0120] In summary, the augmented reality-based intelligent dance teaching feedback method and system proposed in the present invention realizes highly intelligent and personalized dance teaching by innovatively combining multiple advanced technologies. It can not only provide real-time and accurate movement feedback, but also dynamically adjust teaching strategies according to the progress and interests of learners, while continuously optimizing its own performance using group data. This method and system is expected to completely revolutionize the traditional dance teaching model and provide dance enthusiasts with an unprecedented learning experience.

[0121] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, replacement, and improvement made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An intelligent dance teaching feedback method based on augmented reality, characterized in that: include: The acquisition steps include: Acquire the image sequence of the learner's dance movements through a portable camera device; Obtaining preset standard dance movement data; Processing steps include: Based on the image sequence, identifying the learner's dance movements and postures through a movement recognition algorithm; Generating a three-dimensional movement model of the learner according to the dance movement posture; Based on the three-dimensional motion model and the standard dance motion data, construct an augmented reality teaching scene; Output steps include: In the augmented reality teaching scenario, a comparison result between the learner's real-time action and the standard action is presented; Generates real-time feedback on learners' dance movements.

2. The method according to claim 1, characterized in that The method of obtaining the image sequence of the learner's dance movements by the portable camera device specifically includes: A plurality of portable camera devices are provided to collect dance movement images of the learner from the front and side respectively; The image sequences captured by the multiple portable camera devices are synchronously acquired.

3. The method according to claim 1, characterized in that The action recognition algorithm is based on the SOP-GCN framework and includes: Use 2D feature map convolution layer to extract spatial features; Use 3D spatiotemporal convolution blocks to extract spatiotemporal features; Select effective feature channels through channel attention mechanism.

4. The method according to claim 1, characterized in that The step of generating the learner's three-dimensional motion model specifically includes: Inputting the dance action posture into a pre-trained generative adversarial network; A 3D action model of the learner is generated based on the optimization objective of minimizing the JS divergence between the conditional and unconditional discriminator distributions.

5. The method according to claim 1, characterized in that: The steps of constructing an augmented reality teaching scene specifically include: Create virtual stage, virtual mirror and virtual camera modules; combining the three-dimensional motion model of the learner with a virtual stage; Present standard dance moves in a virtual mirror.

6. The method according to claim 1, characterized in that Also includes: Obtain learners’ facial expression data; analyzing the learner's interest based on the facial expression data; The teaching content and difficulty are dynamically adjusted according to the interest analysis results.

7. The method according to claim 1, characterized in that The step of generating real-time feedback information for the learner's dance movements specifically includes: Calculate the deviation between the learner's action and the standard action; Score the learner's actions based on the preset scoring criteria; Generate feedback including corrective action suggestions and scores.

8. The method according to claim 1, characterized in that Also includes: Record learners’ learning time and movement accuracy data; Generate a learning progress report based on the learning time and action accuracy data; According to the learning progress report, the subsequent teaching plan is automatically adjusted.

9. The method according to claim 1, characterized in that: Also includes: Obtain learning data of multiple learners; Based on the learning data of the multiple learners, the action recognition model and the feedback generation strategy are optimized through a machine learning algorithm.

10. The intelligent dance teaching feedback system based on augmented reality is characterized by: include: An image acquisition module, used for acquiring an image sequence of a learner's dance movements through a portable camera device; A motion recognition module, used to recognize the dance movements of the learner through a motion recognition algorithm based on the image sequence; A model generation module, used to generate a three-dimensional movement model of the learner according to the dance movement posture; A scene construction module, used to construct an augmented reality teaching scene based on the three-dimensional action model and preset standard dance action data; A display module, used to present the comparison results between the learner's real-time movements and standard movements in the augmented reality teaching scene; The feedback generation module is used to generate real-time feedback information for learners' dance movements.

Images